ASO-DKELM: Alpine skiing Optimization based Deep Kernel Extreme Learning Machine for Elderly Stroke Detection from EEG Signal

• Published in: Biomedical signal processing and control 2024-02, Vol.88, p.105295, Article 105295
• Main Authors: Nancy, P., Parameswari, M., Sathya Priya, J.
• Format: Article
• Language: English
• Subjects: Alpine skiing optimization; Deep Kernel Extreme Learning Machine; Electroencephalogram; Fast Hartley Transform; Stroke detection
• ISSN: 1746-8094; 1746-8108
• DOI: 10.1016/j.bspc.2023.105295

[Display omitted] •DKELM-AS approach predicts the probability of subjects being affected by stroke.•It classifies stroke affected EEG and normal EEG based on the precursor symptoms.•EEGsignals of normal and stroke patients are used to analyze the proposed method.•Artifacts and noises present in the raw data are eliminated by preprocessing step.•Features are extracted based on frequency ranges using FHT technique. Stroke is considered the third leading cause of mortality followed by cancer and heart disease worldwide. Stroke detection in advance is the only procedure to avoid health risks while it necessitates continuous monitoring and observation of subjects. Many research works already exist in the medical field to detect diseases in advance using machine learning techniques. However, their detection results are limited by several biasing factors such as more training time and imprecise detection. So a novel stroke detection system using the Deep Kernel Extreme Learning Machine based Alpine Skiing (DKELM-AS) approach is proposed to predict the probability of subjects being affected by stroke disease. The electroencephalogram (EEG) signals of normal and stroke patients are collected and utilized to investigate the proposed approach. The raw data are preprocessed and the features are extracted based on frequency ranges using the Fast Hartley Transform (FHT) technique. Finally, the proposed DKELM-AS approach classifies them as stroke-affected EEG or normal EEG based on the precursor symptoms obtained from EEG signals. The experimental results reveal that the proposed DKELM-AS approach achieves an overall accuracy rate of about 95.2% in stroke detection based on the precursor symptoms of EEG signals.